This paper proposes a method to characterize circadian changes in QT intervals for diabetic studies. Although properties of QT intervals for diabetic patients are extensively studied, their circadian changes are not fully understood. Recently, the traditional cosinor method has been utilized for a study examining the relationship between QT circadian changes and the insulin resistance of the diabetic patients. For better characterization of the circadian change in QT intervals of this kind, spline smoothing technique applied to a decimated data set of QT intervals is proposed. New indices named QT circadian transition time (QTCT) and QT circadian transition amplitude (QTCA) associated with the subjects' awakening process are defined to characterize diabetic patients' condition. The method is applied to ten normal and fifteen type 2 diabetic patients. The proposed indices showed significantly lower values for type 2 diabetic patients compared to the control subjects indicating their effectiveness for the characterization.